Today's 802.11 WLANs, which facilitate both 802.11b and 802.11g devices, often experience rate diversity in which multiple nodes attempt to transmit on the shared channel using different speeds. In the presence of rate diversity, the 802.11 MAC protocol DCF (for Distributed Coordination Function) allocates channel capacity in a way that i) degrades aggregate throughputs and ii) leads rational nodes to arrive at undesirable equilibria in which the wireless channel is used inefficiently.
In this talk, we outline the sources of these problems and present practical and effective solutions. We show that traditionally accepted fairness notions such as DCF's frame-based fairness are the root of the problems and argue that an alternate time-based fairness notion should be used instead.
We implement two solutions, TBR (for Time-based Regulator) and DTS (for Distributed Time-based Scheduler), that provide an efficient time-based fairness notion. TBR is a centralized solution running at the AP and requires no modifications to clients nor to DCF. DTS is a distributed, MAC-layer solution that is effective in any environment, including non-cooperative multi-cell environments. Unlike DCF, DTS i) provides equal time shares among contending flows in the presence of rate diversity and varying channel conditions, ii) leads rational nodes to arrive at desirable equilibria and iii) scales well with the number of clients and flows. Our extensive simulation results and preliminary results using our Linux-based implementations show that that both DTS and TBR achieve predicted performance gains (as much as 100% improvement in aggregate throughputs) over DCF.